Instantaneous Assessment of Hedonic Olfactory Perception using Heartbeat Nonlinear Dynamics: A Preliminary Study

Emotional perception to hedonic olfactory stimuli is un- der direct control of the limbic system, whose dynam- ics is known to affect autonomic nervous system activity on cardiovascular control. Mainly due to methodologi- cal limitations, previous investigations failed to uncover specific trends in heartbeat dynamics between ultra short- time (i.e., lasting < 10s), pleasant and unpleasant olfac- tory stimuli. To this extent, in this study we computed in- stantaneous estimates from heartbeat series gathered from 32 healthy subjects (age: 26±2; 16M) undergoing hedo- nic olfactory elicitation. Each subject exhibited a simi- lar olfactory perception threshold, and scored five 5s stim- uli in terms of arousal and valence level using the self- assessment manikin test. We analyzed the heartbeat series using our recently proposed inhomogeneous point-process nonlinear framework, obtaining instantaneous estimates defined in the time (mean and standard deviation), fre- quency (power in the LF and HF bands, as well as LF/HF ratio), and nonlinear/complexity (bispectra, sample and approximate entropy, and Lyapunov exponent) domains. A feature set comprising average estimates within the 5s win- dows was taken as an input of a K-Nearest Neighborhood classification algorithm, whose cross-validation relied on a leave-one-subject-out procedure. Results demonstrate that our framework allows to finely characterize affective olfactory elicitation with an average recognition accuracy of 71.88%. Feature selection highlighted that the most dis- criminating power was contributed by instantaneous LF power, instantaneous Lyapunov exponents, and instanta- neous approximate entropy.